[FR] Should capture the error message after the error occurs in reply() in core.conversation.py

[jkshj21]

Is your proposal related to a problem?

Problem:
Whenever we are running the end to end test using reply function or running the nlu test using intent_detection function, it would be good if it captures the error message and indicate it to the user.

Solution:
when either 4xx (Client error) or 5xx (internal error) error occurs, instead of passing an empty dictionary, we should pass the error data that specifies the error message to quality check the agent behavior.

` try:
response = session_client.detect_intent(request=request)

    except core_exceptions.InternalServerError as err:
        logging.error(
            "---- ERROR --- InternalServerError caught on CX.detect %s", err
        )
        logging.error("text: %s", text)
        logging.error("query_params: %s", query_params)
        logging.error("query_input: %s", query_input)
        return {}

    except core_exceptions.ClientError as err:
        logging.error(
            "---- ERROR ---- ClientError caught on CX.detect %s", err
        )
        template = "An exception of type {0} occurred. \nArguments:\n{1!r}"
        message = template.format(type(err).__name__, err.args)
        logging.error("err name %s", message)

        logging.error("text %s", text)
        logging.error("query_params %s", query_params)
        logging.error("query_input %s", query_input)
        logging.error(traceback.print_exc())
        retries += 1
        if retries < MAX_RETRIES:
            logging.error("retrying")
            return self.reply(send_obj, restart=restart, retries=retries)
        else:
            logging.error("MAX_RETRIES exceeded")
            return {}

`